Face forgery detection method based on multi-region attention mechanism

Abstract

The invention discloses a face forgery detection method based on a multi-region attention mechanism. The method comprises the steps: inputting a to-be-detected face image into a convolutional neural network, and obtaining a shallow-layer feature map, a middle-layer feature map and a deep-layer feature map; performing texture enhancement operation on the shallow feature map to obtain a texture feature map; generating a multi-region attention map for the intermediate layer feature map through a multi-attention mechanism; performing attention pooling on the texture feature map by using a multi-region attention map to obtain local texture features, adding the attention maps, and performing attention pooling on the deep feature map to obtain global features; and the global features and the local texture features are fused and then classified to obtain a face forgery detection result. The method has a plurality of attention regions, and each region can extract mutually independent features to enable the network to pay more attention to local texture information, so that the accuracy of a detection result is improved.

Description

technical field [0001] The invention relates to the technical field of face forgery detection, in particular to a face forgery detection method based on a multi-region attention mechanism. Background technique [0002] Face forgery refers to the use of computer technology to tamper with the face area in media such as images or videos, including identity replacement and expression editing. Face forgery technology can be applied to film and television post-processing. With the tremendous development of deep learning technology in the field of image generation, generative adversarial networks and autoencoders have been applied to the field of face forgery to generate face forgery pictures or videos that are difficult for human eyes to distinguish, such as Deepfakes, FSGAN, FaceShifter, etc. Many face forgery programs can be obtained on the Internet, so that anyone can use a personal computer to synthesize a fake video with a simple study. These fake videos have now appeared wi...

AI Technical Summary

Problems solved by technology

Only a small area of ​​high-quality forged faces is clearly discriminative, so Dang et al. proposed to use the attention mechanism to enhance the accuracy of the face forgery detection network. In computer vision, the attention mechanism is a broad The concept here refers to position-based flexible attention, which multiplies a weight for each position in the feature map; however, this scheme: 1) only uses deep features and ignores texture information; 2) there is only one attention area, Local features are ignored; therefore, the accuracy of the detection results needs to be improved

Images